A Novel Colored Time Petri Net Model for Policy Management and Context Awareness Procedure in MANET

SECAD-Net: Self-Supervised CAD Reconstruction by Learning Sketch-Extrude Operations

“…With the release of several large-scale CAD datasets (e.g., ABC [21], Fusion 360 [50]), SketchGraphs [37]), numerous approaches have explored deep learning to address primitive segmentation/detection [25,56], parametric curve or surface inference from point clouds [17,27,31,40,48] or B-rep models [18,23]. However, by only outputting individual curves or surfaces, these methods lack the CAD modeling operations that are needed to build solid models.…”

Real-Time Multi-Material Reflectance Reconstruction for Large-Scale Scenes Under Uncontrolled Illumination from RGB-D Image Sequences

“…Recently, differentiable rendering became a prevalent tool for appearance reconstruction tasks. More and more differentiable rendering frameworks are published, ranging from general purpose differentiable path-tracers [Li et al, 2018;Nimier-David et al, 2019] over more light-weight solutions [Zhang et al, 2020;Lassner and Zollhofer, 2021] to differentiable rasterization-based renderers [Laine et al, 2020]. With the help of these frameworks, a gradient with respect to scene parameters like illumination, geometry, or reflectance can be calculated for a rendered image, which allows for stochastic gradient descent-based optimization to find parameters matching one or more given input images.…”

“…EC-Net [Yu et al, 2018] identifies edges by encoding local neighborhoods with an underlying PointNet++ [Qi et al, 2017b] and utilizing an edge-aware loss. Similarly, PIE-NET [Wang et al, 2020] uses two PointNet++-like networks and a consecutive non-maximal suppression technique to find the edge and corner points in point clouds. PCPNet [Guerrero et al, 2018] even uses a network inspired by PointNet [Qi et al, 2017a] in conjunction with multi-scale patches as input for estimating normals or curvatures as well as for point classification tasks.…”

“…Other approaches include the neural network based pointwise distance estimation to the next sharp geometric feature [Matveev et al, 2022], or neural network based edge-aware point set consolidation [Yu et al, 2018] and 3D semantic edge detection based on a two-stream fully-convolutional network to jointly perform edge detection and semantic segmentation [Hu et al, 2020]. A further method [Wang et al, 2020] trains two neural networks to classify points into corners and edges based on a PointNet++ like architecture [Qi et al, 2017b]. After a subsequent non-maxima suppression of the classified points and their PointNet++ based clustering, a two-headed PointNet [Qi et al, 2017a] generates the final set of curves.…”

“…Throughout this section, we compare the performance of our work with the performance of several other recent related works for point cloud edge detection: Covariance Analysis (CA) [Bazazian et al, 2015], Feature Edges Estimation (FEE) [Mérigot et al, 2011;Alliez et al, 2022], ECNet [Yu et al, 2018], PIE-NET [Wang et al, 2020], PCPNet [Guerrero et al, 2018],…”